Stop running
your AI agents
in the dark.

Your agents are running in production while you sleep, calling tools and burning tokens you can't see. Jenz catches what they do and stops the runaways.

Vercel AI SDKOpenAI Agent SDKClaude Agent SDKClaude Managed Agents

// problem · the black box

Agents fail silently. While they keep running.

Software fails loud. You get a stack trace, fix it, move on. AI agents don't work that way. They keep reporting ok while something loops in the background, burning tokens nobody's watching.

Once an agent has your tools and a runtime, it can do almost anything. Stack traces won't tell you why it called the same tool 800 times at 3am, or what prompt sent it spiraling. You find out from the bill.

blackboxstatus · ok
tokens: ???tools: ???cost: ???
from the builders

Your agents are running in the dark. Help us flip the lights on.

Building agents is the most fun we've had as developers in years. They keep doing what shouldn't be possible, and breaking in ways nothing else does. We kept wishing something like Jenz existed, and nobody had built it. So we started.

If you're running agents in production, or thinking about it, hop in and tell us what to ship next. The earlier you join, the more of Jenz you get to shape.

— the jenz team

// observability

See every prompt, token, and tool, across every agent and run.

Replay any run, watch tokens climb in real time, and see every tool the agent reaches for. Workspaces keep it all organized, from the moment your cron fires to the final response.

tracerun #4821
run.startsupport-bot1.4s
├─llm.callclaude-3.5-sonnet620ms
├─tool.callgoogle.search820ms
└─tool.callpostgres.query1.1s

Replay any run, end to end

Replay any run, step by step. See every prompt, tool call, and retry your agent made, instead of guessing.

tokenslast 24h
2.4M$128
claude-3.5-sonnet1.4M
gpt-4o768k
gemini-1.5-pro224k

Tokens and cost per LLM

Live token spend across Claude, GPT, Gemini, and whatever else. Catch a runaway before it hits your invoice.

tool callsthis run
github.search234 calls
notion.read_page187 calls
linear.create_issue98 calls
slack.post_message64 calls

Every tool call, logged

See every tool call your agent makes, with args and results. Web search, database, MCP, anything custom, all logged.

agentspaces
production3 agents
└support-botrunning
└research-agentrunning
└summarizeridle

Workspaces for your agents

Keep agents organized by team or project. Each space has its own dashboard, traces, and alerts.

// kill switch

Catch the problems, kill the runaways.

You get an alert the moment something breaks, and you can start, stop, or pause any agent yourself. If one starts looping, spiking tokens, or tripping a watchdog rule, Jenz kills it automatically.

alert
run failed2m ago
support-bot · tool web_search timed out after 30s
sent to

Instant alerts when things break

Slack, email, or webhook the second something breaks. You hear about it before the customer does.

watchdogauto · on
research-agentrunning
summarizerpaused
runaway-cron · killed by jenz
token cap exceeded · stopped in 0.6s

Manual controls and auto-kill

Start, stop, or pause any agent yourself. Set token caps and watchdog rules, and Jenz auto-kills the runaways.

// preview · setup

30-second setup.

coming soon

Setup is one prompt. Pick your SDK, paste it into your AI editor, and Jenz wires itself in before you switch tabs.

prompt5 words

>Set up Jenz with @jenz-ai/vercel-ai.